.
@martin_casado on the paradox that makes model routing an "AI-complete problem":
"Let's imagine you're trying to answer the question, what question does the smartest thing in the universe need to answer? You need the smartest thing in the universe to answer that question."
"The only thing that can answer that question is the smartest model, and then in which case you just give it to the smartest model."
"Clearly the frontier models are getting very jagged. This isn't a core technical capability. It's the decisions of the people that created it decided to make it very good in front end, or very good in 3D, or very good in language. It's not like this one's smarter than this one. It's like this one's very good at this type of thing versus very good at this type of thing."
"We have a lot of specialist models, and a model router will just decide this is better in code, this is better in language, this is better in front end."
"But if you look at model routing and the gains as deployed by application companies, the primary gains tend to be keeping the quality bar high while decreasing the cost."
@a16z